R help - Jan 2014 - Selecting individuals to maximize the correlation of two variables

Hi,
Maybe it is not directly related to R but sine many are statistical experts so I
post it here for help:

I have two variables (say x and y) of length n. Now the cor(x,y) is close to 0.
I need to find the subset in {1,.. n} so that the correlation between x and y
using the subset data is maximized. A trivial choice would be selecting 2
individuals only so that cor(x,y) =1. As the size of the subset increases,
cor(x,y) will go down to 0, but I am assuming the best correlation for each size
of the subsets would not be monotonically decreasing.

Any idea of how to find the solution?

Thanks,
Jing

 		 	   		  
	[[alternative HTML version deleted]]

On Fri, 10 Jan 2014, ?? <mr_tangjing at hotmail.com> writes:
> Hi,
> Maybe it is not directly related to R but sine many are statistical experts
so I post it here for help:
>
> I have two variables (say x and y) of length n. Now the cor(x,y) is close
to 0. I need to find the subset in {1,.. n} so that the correlation between x
and y using the subset data is maximized. A trivial choice would be selecting 2
individuals only so that cor(x,y) =1. As the size of the subset increases,
cor(x,y) will go down to 0, but I am assuming the best correlation for each size
of the subsets would not be monotonically decreasing.
>
> Any idea of how to find the solution?
>
> Thanks,
> Jing
>
>  		 	   		  
Hi Jing,

in chapter 1 of the NMOF manual I discuss a very similar problem; perhaps
it helps you (but it's a draft only...)

http://enricoschumann.net/NMOF.htm#NMOFmanual


-- 
Enrico Schumann
Lucerne, Switzerland
http://enricoschumann.net